Computer Vision Basics with openCV

Import statements

import cv2
import numpy as np

Check your openCV version

cv2.__version__

'4.0.1'

Computer Vision Basics

Load and Image

# Declare a variable with the path to your image
img = cv2.imread('annotations.jpg')

# Load the image with openCV

Hint

Check the image type

# What is the image type? -> type
type(img)

numpy.ndarray

Hint

# Print the value of the pixels of the image
img

array([[[ 50,  56,  55],
        [ 52,  58,  57],
        [ 54,  60,  59],
        ...,
        [ 35,  95, 137],
        [ 27,  54,  91],
        [ 56,  68,  62]],

       [[ 51,  57,  56],
        [ 52,  58,  57],
        [ 54,  60,  59],
        ...,
        [ 28,  80, 120],
        [ 32,  53,  85],
        [ 55,  63,  56]],

       [[ 51,  57,  56],
        [ 52,  58,  57],
        [ 54,  60,  59],
        ...,
        [ 14,  56,  91],
        [ 44,  57,  83],
        [ 56,  57,  47]],

       ...,

       [[177, 166, 158],
        [182, 171, 163],
        [183, 172, 164],
        ...,
        [ 61,  63,  74],
        [ 40,  42,  53],
        [ 48,  50,  60]],

       [[179, 168, 160],
        [181, 170, 162],
        [178, 167, 159],
        ...,
        [ 61,  62,  76],
        [ 37,  39,  50],
        [ 49,  51,  61]],

       [[181, 171, 161],
        [178, 168, 158],
        [172, 162, 152],
        ...,
        [ 60,  61,  75],
        [ 30,  32,  43],
        [ 42,  44,  55]]], dtype=uint8)

Hint

Show the image using openCV

# Lets now see the image! dont forget to call destroyAllWindows!
cv2.imshow('My Image', img)
cv2.waitKey(0)
cv2.destroyAllWindows()
cv2.waitKey(1)

-1

Hint

Now lets display unsing matplotlib

# Import matplotlib
import matplotlib.pyplot as plt
%matplotlib inline

# Choose a pyplot figsize to make images fit nicely into the notebook
plt.figure( figsize = (20,15))


# Now display the image with plt
plt.imshow(img)

<matplotlib.image.AxesImage at 0x16f2ba7a8b0>

Lets fix the colors

# Does your image look ok? are the colors the right ones?, if not, lets fix it!
# Remember that openCV will load images using BGR rather than RGB
# Change the image to RGB and diplay it again with plt
rgb_img=cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
plt.figure( figsize = (20,15))
plt.imshow(rgb_img)

<matplotlib.image.AxesImage at 0x16f2c2ee970>

Hint

What is the shape of the image?

# Check the image shape
img.shape

(576, 768, 3)

Lets make it grayscale

# Load the the image in grayscale, what are the 2 ways you can do that?
grayscale_img = cv2.imread('annotations.jpg', cv2.IMREAD_GRAYSCALE)

# Now load an image in color and then transform it to grayscale
grayscale_img = cv2.imread('annotations.jpg', cv2.IMREAD_GRAYSCALE)

What is the shape now?

# Check the grayscale shape
grayscale_img

array([[ 55,  57,  59, ..., 101,  62,  65],
       [ 56,  57,  59, ...,  86,  60,  60],
       [ 56,  57,  59, ...,  62,  63,  54],
       ...,
       [165, 170, 171, ...,  66,  45,  53],
       [167, 169, 166, ...,  66,  42,  54],
       [169, 166, 160, ...,  65,  35,  47]], dtype=uint8)

How does it look like?

# Display the grayscale image with matplotlib, how does it look?
plt.figure( figsize = (20,15))
plt.imshow(grayscale_img)

<matplotlib.image.AxesImage at 0x16f2c3a5bb0>

We need to indicate the colormap

# Display the grayscale image with matplotlib, make sure to include the colormap so it really is grayscale
plt.figure( figsize = (20,15))
plt.imshow(grayscale_img, cmap='gray')

<matplotlib.image.AxesImage at 0x16f2c40bd60>

Cropping an image

# Lets now crop a Region of Interest of an image, load and image and crop different parts of it
cropped_img = rgb_img[0:250, 0:600]

# Save to disk the cropped areas
cv2.imwrite('sky_crop.jpg',cropped_img )

True

To edit an image making a copy of it is always a good idea!

# Load an image and then make a copy of it
loaded_img = cv2.imread('sky_crop.jpg')
plt.figure( figsize = (20,15))
plt.imshow(loaded_img)

<matplotlib.image.AxesImage at 0x16f2d487fa0>

Hint

Annotating images

Rectangles

# Draw green rectangle on top of a region of interest in your image
green =(0,255,0)

Hint

Adding text to images

# Make a copy of your image and then write some text on top of the image
clone = rgb_img.copy()
clone

array([[[ 55,  56,  50],
        [ 57,  58,  52],
        [ 59,  60,  54],
        ...,
        [137,  95,  35],
        [ 91,  54,  27],
        [ 62,  68,  56]],

       [[ 56,  57,  51],
        [ 57,  58,  52],
        [ 59,  60,  54],
        ...,
        [120,  80,  28],
        [ 85,  53,  32],
        [ 56,  63,  55]],

       [[ 56,  57,  51],
        [ 57,  58,  52],
        [ 59,  60,  54],
        ...,
        [ 91,  56,  14],
        [ 83,  57,  44],
        [ 47,  57,  56]],

       ...,

       [[158, 166, 177],
        [163, 171, 182],
        [164, 172, 183],
        ...,
        [ 74,  63,  61],
        [ 53,  42,  40],
        [ 60,  50,  48]],

       [[160, 168, 179],
        [162, 170, 181],
        [159, 167, 178],
        ...,
        [ 76,  62,  61],
        [ 50,  39,  37],
        [ 61,  51,  49]],

       [[161, 171, 181],
        [158, 168, 178],
        [152, 162, 172],
        ...,
        [ 75,  61,  60],
        [ 43,  32,  30],
        [ 55,  44,  42]]], dtype=uint8)

array([[[ 55,  56,  50],
        [ 57,  58,  52],
        [ 59,  60,  54],
        ...,
        [137,  95,  35],
        [ 91,  54,  27],
        [ 62,  68,  56]],

       [[ 56,  57,  51],
        [ 57,  58,  52],
        [ 59,  60,  54],
        ...,
        [120,  80,  28],
        [ 85,  53,  32],
        [ 56,  63,  55]],

       [[ 56,  57,  51],
        [ 57,  58,  52],
        [ 59,  60,  54],
        ...,
        [ 91,  56,  14],
        [ 83,  57,  44],
        [ 47,  57,  56]],

       ...,

       [[158, 166, 177],
        [163, 171, 182],
        [164, 172, 183],
        ...,
        [ 74,  63,  61],
        [ 53,  42,  40],
        [ 60,  50,  48]],

       [[160, 168, 179],
        [162, 170, 181],
        [159, 167, 178],
        ...,
        [ 76,  62,  61],
        [ 50,  39,  37],
        [ 61,  51,  49]],

       [[161, 171, 181],
        [158, 168, 178],
        [152, 162, 172],
        ...,
        [ 75,  61,  60],
        [ 43,  32,  30],
        [ 55,  44,  42]]], dtype=uint8)

Hint

Now that you know how to load annotate and save images lets start manually annotating some images

# Load as many images as you like and start annotating objects in the image, 
# Before annotating the images make a copy of the image so we keep the original without changes
# Use the format proposed in the example above
cv2.rectangle(clone,(100,200),(400,500), green, 3)
plt.figure( figsize = (20,15))
plt.imshow(clone)

<matplotlib.image.AxesImage at 0x16f2d638340>

# Display the images that you have annotated above, both versions, the original one and the annotated one
plt.figure( figsize = (20,15))
plt.imshow(rgb_img)

<matplotlib.image.AxesImage at 0x16f2c380250>

Mini Projects

With all that you accomplished to do above you are now ready to do your first mini project!, you might need to learn some more new things to accomplish some of the harder levels below, it is very important to get used to the openCV docs

Meme creator

Level 1

1. Create a basic Meme creator with one image and a short line of text, using any openCV font

Level 2

1. Use the classic meme font (impact)

Level 3

1. Improve it by adding 2 to 3 lines with the text on the top of the image and the bottom of the image
2. Make the text in white and with black contours

Level 4

1. Make the text center automatically Hint
2. If the text is too long it should wrap the text in as many lines as needed

QR Reader

Level 1

1. Researn on the openCV docs and learn about QRCodeDetector
2. Grab any QR code from google images for example this one: or any other.
3. Detect the code using QRCodeDetector
4. Print QR detected if a QR has been detected in the image

Level 2

1. Decode and print the decoded text
2. Draw a rectangle around the QR code that has been detected and show it on screen
3. Save the QR region of the image to disk